Big Data Business Analytics

Analytical Projects

For several businesses, I've worked on analytical projects. Below are some demos and previews of my work.

• Quantitative Customer Journey Predictive Analytics

• Display Banner Rotation Optimizer

• Behavioural Customer Trend Anomaly Detection

• GasShipping Optimization Model

• Value At Risk Energy Trading Model

• Commercial Spot Frequency Value Calculator

• Movie Scheduler

• Data Point Editor

• PurpleSpy++ Computer Security Monitor

Quantitative Customer Journey Predictive Analytics

Programming language

Teradata / SAS / MicroStrategy / Celonis Process Mining

Operating system

Unix

Release dates

2015 (MVP), 2016 (MicroStrategy Report), 2017 (Predictive Module), 2018 (Process Mining Module)

Type

1-on-1 Marketing forecasting model

Description

The Quantitative Customer Journey Predictive Analytics big data tool developed for a major financial institution in The Netherlands, enables the 1-on-1 forecasting of retail marketing campaign effectiveness and smart target audience selection. The big data tool running on Teradata platform daily tracks customers and prospects end to end in their cross channel journeys for buying, using and retaining financial products like savings accounts, loans, residential mortgages and investments. This custom built Customer Journey suite runs nightly in order to provide next-day business intelligence insights for retail marketing campaign effectiveness using MicroStrategy dashboards. Behind the scenes the forecasting module implemented in SAS Enterprise Miner automatically optimizes running campaigns for net journey effect, such that campaigns will be automatically adjusted to the most promissing audience with regard to sales and service. Finally, the opportunity insight component discovers and signals previously hidden marketing opportunities across channels and products.

Display Banner Rotation Optimizer

Programming language

Teradata, MicroStrategy

Operating system

Unix

Release dates

2017

Type

Display Banner 1-on-1 Reach Planning Model

Description

The Display Banner Rotation Optimizer consists of a first customer response eliciting mechanism based on group similarity characteristics of the client. By optimally rotating random display banners with a higher than average probability of click-through for the comparison group, online customer interactivity for retail clients of a major Dutch financial institution was increased. The statistical model behind the banner planning system was quite simple in mathematical terms, but very valuable for business value generation as no longer the default banners were served to customers. The improved approach of banner rotation resulted in a higher reach and better response on the on-site bannering channel. The results were monitored in a MicroStrategy dashboard of the marketing department.

Behavioral Customer Trend Anomaly Detection

Programming language

Teradata / SAS

Operating system

Unix

Release dates

2017

Type

Trend Anomaly Detection model

Description

The Behavioral Customer Trend Anomaly Detection developed for a major financial institution in The Netherlands, automatically tracks any significant changes in many customer behavioural attributes like cross sell, channel usage and financial status. The big data tool runs on a monthly basis on the SAS Unix platform, producing actionable signals marking customers whose behavior trend has changed above/below a certain personalized threshold. Allowing the financial institution to monitor millions of customers on changing patterns in behavior and attributes, such as for purposes of acquisition, retainment or churn. The software monthly-refreshed results are easily imported into other tooling such as marketing automation and fraud detection monitoring systems.

GasShipping Optimization Model

Programming language

MATLAB

Operating system

Windows Server 2003

Release date

2008

Type

Financial computation model

Description

The hourly gas shipping allocations for the energy firm needed to comply with the regulations of the Network Supervisor. Also the allocated capacity may not differ too much from the real used capacity, otherwise penalties would be given to the gas shipper. A VBA-model, with lots of sheets and formulas, was turned into an easy to maintain automation script. Computation time was improved significantly and several gas portfolio's could be calculated and optimized at the same time. Computational results were both exported to a spreadsheet as well as understandably presented in graphs and charts for management. With the newly developed software, more savings could be achieved on avoiding penalties and by better pricing.

Value At Risk Energy Trading Model

Programming language

MATLAB

Operating system

Windows Server 2003

Release date

2006

Type

Financial computation model

Description

Using the mathematical programming language a financial model was turned into a automation script. With the Energy Trading Model the daily financial risk positions for the commodity portfolio (Energy, Gas, Oil) could be calculated. The thousands of Value-at-Risk positions were calculated in mere seconds, and when the model was done a user interpretable graph was displayed to give an overview of total risk. I.e. green is good low risk, red is bad high risk.

Commercial Spot Frequency Value Calculator

Programming language

Java

Operating system

Windows XP

Release date

2006

Type

Econometric optimization model

Description

Software written for my Master's Thesis that implemented a mathematical model for computing the Reach and Frequency for television spot time slot programming. The program loaded a commercial break schedule as well as viewer probability matrices (e.g. retrieved from a rating agency) and then outputted the best scheduling chances in XML.

Movie Scheduler

Programming language

C++ / VBA / Java

Operating system

Windows XP

Release date

2005

Type

Econometric optimization model

Description

Software written for my Bachelor's Thesis that implemented a mathematical model for allocating cinema movies to time slots in order to maximize ticket sales. The algorithm included the use of ILOG CPLEX software for column generation and branch and bound Lagrangian search. The output was written to XML and loaded back into a spreadsheet for management using VBA-code (1^st screenshot). In a second iteration I improved the Graphical User Interface (GUI) to a Java interface (2^nd screenshot).

Data Point Editor

Programming language

Java (Niagara Baja Framework)

Operating system

Java embedded systems

Release date

2004

Type

Automation and productivity software

Description

The Data Point Editor is a GUI-tool for automating climate control embedded systems programming/configuration tasks. With the Data Point Editor building automation project manager can easily connect climate control devices like heaters, vents, lights and motion sensors, with software logic and room switches. Team productivity increased, projects were finished earlier and more business customers could be served. Over the years other colleague improved the Data Point Editor further.

PurpleSpy++ Computer Security Monitor

Programming language

Visual Basic

Operating system

Windows 95

Release date

2000

Type

Security software

Description

The application was designed to protect the computer lab environment for misbehaving high-school students. PurpleSpy++ runs on the background and monitors unusual activity by the user. Observations are reported in a log for the IT-staff. In a sense, the program made it possible to have Guest Users at your Windows 95 computer, without them damaging or removing your files.